POVERTY & EQUITY NOTES JUNE 2020 · NUMBER 24 Health and Distributional Effects Taxing Sugar-Sweetened Beverages: the case of Kazakhstan Alan Fuchs, Kate Mandeville, Ana Cristina Alonso-Soria Consumption of sugar-sweetened beverages (SSBs) has been linked to a range of noncommunicable diseases (NCDs), including diabetes, obesity, cardiovascular disease, and over 12 types of cancer (Singh, et al. 2015). Increasingly, governments around the world are taxing SSBs to curb sugar consumption. However, especially in low and middle-income countries, concern is growing over the apparent regressive character of consumer taxes. This note contributes to the literature on the effect of taxes on unhealthy products. The extended cost-benefit analysis (ECBA) methodology is applied to assess distributional effects of an increase of SSB taxes on household expenditures, out-of-pocket (OOP) medical expenses, and productivity in Kazakhstan. Results suggest that the long-term net income effect of an increase in SSB taxes is progressive: lower-income deciles benefit relatively more than higher-income deciles. Mounting evidence links consumption of SSBs to a Over 85 percent of premature deaths from NCDs occur variety of negative health effects, as well as economic in low- and middle-income countries. Promotion of costs from health care expenditures, forgone healthy diets to control NCDs worldwide target SSBs. The productivity, disability, and premature death. SSBs— achievement of the Sustainable Development Goal (SDG) drinks containing added caloric sweeteners, such as target of a one-third reduction in premature deaths from sucrose, high-fructose corn syrup, or fruit juice NCDs by 2030 requires immediate actions to reduce concentrates—include carbonates, fruit beverages, sports dietary risks, especially sugar consumption. beverages, energy and vitamin water beverages, sweetened iced tea, and lemonade (WCRF 2018). SSB As of 2019, more than 37 countries had implemented consumption has been linked to ischemic heart disease, tax policies on sugary drinks, including Ecuador, India, hypertension, type 2 diabetes (GBD 2018; WCRF 2018), Ireland, Mexico, Peru, the Philippines, South Africa, and hypertension (Xi, et al. 2015), long-term weight gain (Hu Thailand (Cawley, et al. 2019). While price increases from 2013), tooth decay, and 12 types of cancer, including taxes on unhealthy products represent a large short-term esophageal, colon, pancreatic, breast, uterine, kidney, and burden on low-income households, studies have also gall bladder (Singh, et al. 2015). shown that the largest longer-run economic and health benefits also accrue to low-income consumers because of Taxes to increase the prices of SSBs by at least 20 their stronger response to price changes (Sassi, et al. percent is one of the most effective tools for reducing 2018). 1 obesity rates and related NCDs, according to the World Health Organization (WHO 2017a). Epidemiological This note summarizes the first use of the extended models indicate that taxing SSBs by sugar content could cost-benefit analysis (ECBA) to examine the income result in 200 million pounds (90.7 million kilograms) of distribution effects of taxes on SSBs. 2 The main weight reduction worldwide (Grummon, et al. 2019). outcome of interest is the net effect of taxing SSBs on 1 Allcott, Lockwood, and Taubinsky (2019) also developed a novel more price-responsive, low-income consumers, bigger tax model incorporating internality (misperception of long-term costs) and increases result in bigger welfare gains. externality costs from the consumption of SSBs, to show that taxing 2 The ECBA has been applied and adapted to the analysis of these goods is welfare-enhancing. Similarly, they also show that for tobacco taxation in several countries by Fuchs and Meneses household income via 3 channels: (a) larger household YLL lost across deciles proportionally with the number of budget expenditure on SSBs, (b) savings in out-of-pocket households that consume SSBs in each decile. (OOP) spending on health care because of lower disease Net Income Gains: We estimate net income gains by incidence, and (c) higher labor income resulting from decile by adding the price effects, the savings in OOP increased life-expectancy. spending, and the gains in years of productive life. Methodology Data Our ECBA model (Equation 1) calculates the aggregate We use data on household consumption and effect of taxing SSBs. 3 expenditures on SSBs in Kazakhstan from 11 waves of the Household Budget Survey (HBS), from 2007 to Equation 1: Aggregate Effect on SSB Taxes 2017. The survey includes disaggregated consumption Net Change in Change in Change in information on a range of products, including carbonated Income = SSB + OOP spending + years of drinks and fruit juices. 4 We obtain unit prices by dividing Effect expenditure on health care productive life (A) (B) (C) total quantities consumed by the amounts paid, a common practice in the literature. 5 We use HBS information on the ages of family members, gender, SSB Expenditure: We estimate the effect of SSB price educational attainment, and urban versus rural residence increases on household expenditures using the change in to control for household characteristics. prices resulting from the tax, the price elasticity of SSBs by income decile, and the share of household spending on Around 60 percent of households consume SSBs, and SSBs. The model assumes a price increase of 20 percent. this share does not vary significantly with income. We aggregate changes in household expenditures at the Annual average household consumption of SSBs in liters decile level. is higher among higher-income households (28.7 versus 21.9 total annual liters in the top versus the bottom decile OOP Spending: Based on a simple static model, we in 2017). In 2018, total consumption of carbonated drinks estimate the long-term change in OOP spending on and juices in Kazakhstan was approximately 1,206.3 health care that would result from a reduction in the million liters (Euromonitor International 2019). 6 HBS consumption of SSBs. We distribute the cost of treating estimates show that households in the richest deciles pay diseases related to SSB consumption across income a slightly higher average unit price than households in the deciles according to the share of households that poorest deciles, presumably because of quality consume SSBs in each decile. preferences. Years of Productive Life: We derive the effects of The proportion of household expenditures on SSBs reducing SSB consumption on labor income from the has increased over time, from under 1 percent of total reduction in workers’ years of life lost (YLL) because of household expenditures to a little over 1 percent, on premature mortality (deaths before age 70). We estimate average (Figure 1). The share of expenditures on SSBs out the increase in years of productive life by distributing the of total expenditures is lower among richer households. (2017a, 2017b, 2018), Fuchs and Gonzalez Icaza (2020), etc., based on distinction between artificially and non-artificially sweetened Pichón-Riviere et al. (2014) and Verguet et al. (2015). beverages. 3 The model assumes that the reduction in SSB consumption has an 5 An important caveat associated with the elasticity estimates derived immediate effect on health and, hence, on employment-related from unit prices is that the lack of direct price information may result in income. biased estimates, because quality differentials are not directly 4 Our analysis aggregates carbonated drinks and fruit juices, as both identifiable from unit prices. types of products are broadly SSBs. The survey also does not allow 6 Average unit prices from HBS data differ from market price data because of recall error, brand and quality effects, and survey design. June 2020 · Number 24 2 Figure 1: Share of Expenditures on SSBs in Approximately 8.2 percent of total deaths could be Total Household Consumption attributable to SSB consumption (Table 1). We 1.40% collected data from the Global Burden of Disease (GBD) 1.20% database on SSB-related diseases, including mortality 1.00% associated with diabetes mellitus type 2, ischemic heart 0.80% disease, hypertensive heart disease, and stroke in Kazakhstan. 8 And we multiplied the total mortality in 0.60% 2016 for each disease by the corresponding population 0.40% attributable fraction (PAF)—the contribution of a risk 0.20% factor to a disease or a death—associated with each type 0.00% of SSB-related disease. The incidence of diseases 1 2 3 4 5 6 7 8 9 10 Decile associated with SSB consumption (number of newly 2010 2014 2017 diagnosed cases) reached 2.16 million cases in 2016. The largest share of cases corresponded to dental care and Source: Estimates based on the Household Budget Survey obesity. 9 The average price elasticity of SSBs is −0.70, in line with the results found in the literature. We estimated Table 1: Deaths by Gender, Diseases Related to price elasticities using average annual unit prices. After Consumption of SSBs, Kazakhstan, 2016 Rel. PAF % Death: Death: Total % Total eliminating quantity and unit value outliers outside 3 women men death Risk standard deviations, we estimated elasticities for the total Diabetes 1.18 10.4 59 41 100 10 population using a simple equation. 7 An average price Ischemic 1.16 9.3 1,803 1,763 3,566 9 elasticity of demand of −0.70 implies that a tax that heart disease increases the price of SSBs by 20 percent would result in Stroke 1.1 6.0 639 536 1,175 6 a 14 percent drop in quantity demanded. We evaluated 3 Hypertensive 1.12 7.1 62 50 112 7 elasticity scenarios: medium-bound elasticity estimate heart disease with average of -0.70; and lower- and upper-bound Total 2,562 2,390 4,953 8 elasticity scenarios of 0.20 above and 0.20 below the Source: Estimates based on Global Burden of Disease data, the 2016 medium-bound, respectively. Household Budget Survey, and the relative risk estimates review in Shim et al. 2019. Figure 2: Price Elasticities of SSBs, by Decile 0.00 Annual OOP expenditures attributable to the 1 2 3 4 5 6 7 8 9 10 management of diseases related to SSB consumption add up to US$10.7 million, with close to 49 percent of the Price elasticity -0.50 costs corresponding to treatment of cardiovascular disease and the rest to endocrine and metabolic diseases -1.00 (including diabetes). 10 We used data on total medical costs in 2016 from the National Health Accounts of Kazakhstan. -1.50 Income decile Expenditures by disease were not available; thus, we substituted the share of medical expenditures by disease Low Medium Upper group from Belarus, a country with comparable OOP Source: Estimates based on data of Household Budget Survey 2007–17. 7 lnQ_id=β_0+β_1 lnP*D_i+β_3 X_id+μ_id : where Qid represents the attributable fraction (PAF). Thus, we could be underestimating the total quantity of SSBs consumed per year by household i in decile d, benefits of the tax in health and welfare of the population. measured in liters; P the average price of SSB liter; D_i the consumption 9 We used 2016 because this was the latest year with available decile of household i; and X_id a vector of household characteristics information on total medical spending. (urban versus rural, household size, ages of household members, 10 Lack of disaggregated data on treatment costs for specific diseases educational level, and gender of the household head). could potentially result in overestimation of the total costs, as 8 We did not include data on cancer mortality because of lack of groupings contain non-SSB–related diseases. However, this effect suitable parameters to estimate the corresponding population June 2020 · Number 24 3 expenditures as a share of total health care costs (35.80 Incorporating the price elasticities of demand reduces percent in 2015) and a similar disease burden. 11 We the negative impacts of higher SSB expenditures. The estimated the share of total medical expenditures in medium- and lower-bound elasticity scenarios result in Kazakhstan for 2 groups of diseases: (a) endocrine and negative income shocks across all deciles. Only the upper- metabolic diseases (including diabetes) and (b) bound elasticity scenario results in positive income gains cardiovascular disease. We then multiplied total costs per for deciles 1 to 9. Overall, the lower-, medium-, and disease group by the PAF to estimate costs corresponding upper-bound elasticity scenarios show progressivity, that to diseases linked to SSB consumption (Table 2). is, poorer households show relatively smaller income loses (or relatively larger gains) compared with richer Table 2: Out-of-Pocket Expenditures on the Treatment of Diseases Related to the Consumption of SSBs in households. Kazakhstan (2016) Total out-of-pocket Total out-of-pocket Savings in OOP expenses are proportionally higher for expenditures expenditures (2011 lower-income deciles, making the tax effect (tenge) PPP$) progressive. Income gains from reduction in OPP Cardiovascular disease T 1,797,629,151 $21,499,758 spending are positive, but small, across scenarios. Income Endocrine and metabolic T 1,721,707,343 $20,591,728 diseases (including diabetes) gains from the change in years of productive life are Total T 3,519,336,494 $42,091,486 positive across all income deciles, but negligible. Source: Estimates based on National Health Accounts of Kazakhstan Wealthier households receive higher income gains from 2016, the Household Budget Survey 2016, Kontsevaya et al. (2018), and working life years as income gains per year of working life relative risks estimates review of Shim et al. 2019. are higher for them. Approximately 8.1 percent of total Years of life lost In the long run, lower-income deciles benefit more (YLL) are due to diseases associated with the SSB than higher-income deciles from the tax. The net consumption. We estimated the YLL by multiplying the income effect is negative for all income deciles under PAF of each disease associated with SSB consumption by lower and medium-bound elasticity scenarios, but the total YLL-by-disease data obtained from the Global positive for the upper-bound, long-term elasticity Burden of Disease database. We interpret the YLL as a scenario. proxy for income forgone because of worker premature mortality. Figure 3: Total Income Effect: Direct and Indirect Effects of Sugary Beverages Taxes 0.060 Net Income Results 0.040 Income Gains (%) 0.020 With no behavioral changes—assuming a complete 0.000 pass-through scenario—increases in the price of SSBs -0.020 1 2 3 4 5 6 7 8 9 10 are completely passed on to consumers spending in -0.040 SSBs, affecting lower-income deciles relatively more. -0.060 These households spend a larger proportion of their -0.080 incomes on consumption of food and beverages. The -0.100 Income deciles immediate effect of a 20 percent price shock would be a Low-bound elasticity Medium-bound elasticity larger share of the household budget allocated to Upper bound elasticity purchasing SSBs (that is, a negative income effect) among all deciles. Source: Estimates based on Household Budget Survey 2007–17. might be offset by the exclusion of treatment costs for other, related compared with 9.5 percent, 61 percent, and 25.2 percent in Belarus, diseases, including various types of cancer. respectively (WHO 2016). 11 The prevalence of diabetes, overweight, and obesity in Kazakhstan was 11.5 percent, 58.7 percent, and 23.5 percent of the population in 2016, June 2020 · Number 24 4 These results add to substantial other benefits of Among the limitations of this approach, is the assumption taxing SSBs. These include additional tax revenue that of immediate (that is, zero discount rate) health-related governments could redistribute back to the poor in public income benefits. The coverage of diseases and treatment goods, including alternatives to SSBs, such as clean water. costs incorporated into the model is incomplete due to Other benefits of taxes targeted to reduce SSB data limitations. This includes certain cancers associated consumption include substantial improvements in health, with SSB consumption. quality of life, and longevity. More research is necessary to understand the SSB Despite lack of important data and other limitations, taxation effects on consumption and health. As more the results of the ECBA model are in line with results developing countries tax SSBs, the benefits of taxation in from other models, such as the behavioral redistributive more resource-constrained environments should model of Allcott, Lockwood, and Taubinsky (2019). become clear. ABOUT THE AUTHORS Alan Fuchs is a Senior Economist at the World Bank’s Poverty and Equity Global Practice. Kate Mandeville is a Senior Health Specialist at the World Bank’s Health, Nutrition and Population Global Practice. Ana Cristina Alonso-Soria is a Consultant at the World Bank’s Finance, Competitiveness and Innovation Global Practice. This note series is intended to summarize good practices and key policy findings on Poverty-related topics. The views expressed in the notes are those of the authors and do not necessarily reflect those of the World Bank, its board or its member countries. Available for download at the World Bank Publications, Documents & Reports site. June 2020 · Number 24 5